Unitree IPO shows a real hardware business, but the humanoid case is still early

Unitree Robotics filed to raise ¥4.2 billion — roughly $610 million — on Shanghai's STAR Market last week, and the numbers in the filing read like a venture capitalist's fever dream. Revenue grew 335% to ¥1.71 billion in 2025. Net profit surged 674%. The company posted a 60.3% gross margin on overall sales. And in the nine months through September, adjusted net profit margin hit 36.88%. Unitree is not losing money on a dream. It is making money on hardware.

The underwriter is CITIC Securities. The filing was accepted by the Shanghai Stock Exchange on March 20, 2026.

But here's the part that doesn't fit the narrative: the filing also contains Unitree's own breakdown of what its humanoid robots are actually doing in the world. Fifty to seventy percent of Unitree's humanoid "industry application" revenue — the commercial category that excludes research and education — comes from enterprise reception and tour-guide use. Corporate lobbies. Trade show booths. Visitor centers. As Reuters reported, the company's prospectus acknowledged that "real-world factory deployment remains limited for now."

The world's most productive robotics company is mostly in the lobby-greeting business.

That is the tension Unitree's IPO has landed on the table, and it is a genuine one. The numbers — 5,500 humanoid units shipped in 2025, 32.4% global market share, 62.9% gross margin on humanoids — are real and they are impressive by any standard in this industry. Per the Gasgoo analysis of the filing, Unitree outshipped overseas rivals Figure AI and Agility Robotics — each at roughly 150 units — by a factor of more than 30. Tesla has not publicly sold a single Optimus. Unitree did this by building vertically: actuators, sensors, embedded control systems, all in-house. The margin follows from that.

Wang Xingxing, Unitree's founder, built the first XDog — a quadruped robot — as a master's project at Shanghai University. He founded Unitree in 2016 in a 50-square-meter office in Hangzhou with roughly $12,000 in prize money and a $300,000 angel check. Nine years later he was the only post-1990s executive invited to sit front row at Xi Jinping's private sector symposium alongside Ren Zhengfei of Huawei, Jack Ma, and BYD's Wang Chuanfu. As the South China Morning Post documented, Wang holds 23.8% direct equity but controls 68.8% of voting rights through dual-class shares — a structure that gives him near-absolute authority over the company he's bringing public.

His stated philosophy is mass-market robotics: robots as platforms, not premium hardware. The G1 humanoid starts at $16,000. The R1 costs $5,900. As Robot Report noted at the May 2024 G1 launch, the comparable Boston Dynamics Spot quadruped costs $74,500. The price gap is not accident — it is the strategy.

But price and volume do not answer the question of what all those robots are for. The factory automation thesis — humanoid robots doing real labor on production lines — is where the trillion-dollar opportunity lives. Unitree's own prospectus suggests the factory revolution has not arrived yet. The reception thesis has.

Unitree has acknowledged the gap. In February 2026 it released a video of G1 humanoids assembling robot components at its own manufacturing facility using a new model called UnifoLM-X1-0. The video ran at double speed. The facility was Unitree's own — an environment engineers control. The distinction that matters: a proprietary factory is not an unfamiliar one. Wang himself has said the "ChatGPT moment" for robotics arrives when a robot can complete 80% of tasks in 80% of unfamiliar environments. He is not claiming that moment has arrived.

The competitive picture adds another layer of complexity to the #1 claim. An Omdia report published in January 2026 ranked Shanghai-based AgiBot first, with more than 5,100 units shipped and 39% global share, putting Unitree second at 4,200 units. As Caixin Global reported, Unitree disputed those figures publicly, claiming 5,500 units and contesting AgiBot's methodology. By the time of the March IPO filing, Unitree's prospectus claimed global leadership and cited industry data — data that remains unverified by independent analysts. Both companies are likely in the 4,200-to-5,500 range. The "global #1" is contested.

The valuation math is where the IPO story gets complicated. Per Caixin Global, Unitree's June 2025 private funding round valued the company at ¥12.7 billion. The IPO seeks ¥4.2 billion. If the raise is at the midpoint, the implied market cap sits around ¥42 billion — roughly 3.3 times the June valuation nine months prior. That math only works if the humanoid deployment curve bends sharply upward and Unitree holds its position. It also requires believing the tour-guide era is a launchpad, not a destination.

The shift to the lower-priced G1 has already compressed margins. Reuters noted the move into the G1 "trimmed gross margin," and humanoid gross margin of 62.9% reflects the mix before the full consumer pricing push. Unitree's overall gross margin improved from 44.18% in 2022 to 59.45% in the first three quarters of 2025 — a real trajectory — but the direction of that curve depends heavily on product mix. R&D spending as a percentage of revenue has fallen from 31.4% in 2023 to 7.7% in 2025. Absolute R&D spend is still growing — roughly ¥90 million annualized — but the ratio decline will be a feature of every bear case.

Investors in the June 2025 round include Meituan at 9.6%, HongShan Capital (formerly Sequoia China) at 7.1%, and Matrix Partners China at 5.5%, per the South China Morning Post. Tencent, Alibaba, Ant Group, Xiaomi, ByteDance, BYD, and Geely have also invested, directly or indirectly — a cap table that reads like a who-is-who of Chinese tech. The IPO is partly a liquidity event for that group. Wang keeps control.

The macro tailwind is real. Nearly 90% of all humanoid robots sold globally in 2025 were Chinese, according to industry data cited across multiple filings. China's 14th Five-Year Plan lists embodied AI as a strategic priority alongside quantum computing, 6G, and brain-computer interfaces. The government wants a domestic humanoid industry and is funding it accordingly — tax incentives, state-backed capital, procurement preference. The EV playbook is not subtle as an analogy, and Unitree is not trying to hide it.

The IPO is a real hardware company asking real money for a bet on an early-stage revolution. The profit exists. The gross margins exist. The unit volumes exist. And the tour-guide era — however genuinely useful it is as a business — is not the factory floor transformation that justifies a 3.3-times re-rating in nine months. The STAR Market will decide what that gap is worth.